flowent 0.3.0 → 0.3.1

package/backend/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "flowent"
-version = "0.3.0"
+version = "0.3.1"
 description = "A workflow orchestration platform for multi-agent collaboration"
 readme = "README.md"
 authors = [

package/backend/src/flowent/api_models.py

@@ -19,6 +19,18 @@ class ProviderModelsResponse(BaseModel):
     models: list[str]
 
 
+class ProviderModelsFailureResponse(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    code: Literal[
+        "connection_failed",
+        "access_denied",
+        "rate_limited",
+        "provider_unavailable",
+        "request_failed",
+    ]
+
+
 class WorkspaceMessagesRequest(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
@@ -35,8 +47,8 @@ class WorkspaceMessageEditRequest(BaseModel):
 class WorkspaceMessageEditResponse(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
+    is_responding: bool = False
     messages: list[StoredMessage]
-    run_id: str | None = None
 
 
 class WorkspaceRespondRequest(BaseModel):
@@ -46,16 +58,9 @@ class WorkspaceRespondRequest(BaseModel):
     message_id: str | None = None
 
 
-class WorkspaceRunResponse(BaseModel):
-    model_config = ConfigDict(extra="forbid")
-
-    run_id: str
-
-
 class WorkspaceClearResponse(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
-    active_run_id: str | None = None
     messages: list[StoredMessage]
     usage_info: TokenUsageInfo | None = None
 

package/backend/src/flowent/llm.py

@@ -85,6 +85,20 @@ class LLMStreamError(RuntimeError):
     pass
 
 
+class ProviderModelFetchFailure(StrEnum):
+    CONNECTION_FAILED = "connection_failed"
+    ACCESS_DENIED = "access_denied"
+    RATE_LIMITED = "rate_limited"
+    PROVIDER_UNAVAILABLE = "provider_unavailable"
+    REQUEST_FAILED = "request_failed"
+
+
+class ProviderModelFetchError(RuntimeError):
+    def __init__(self, failure: ProviderModelFetchFailure) -> None:
+        super().__init__(failure.value)
+        self.failure = failure
+
+
 async def wait_before_llm_retry(attempt_number: int) -> None:
     await asyncio.sleep(LLM_RETRY_BASE_DELAY_SECONDS * attempt_number)
 
@@ -248,6 +262,31 @@ def unique_model_names(provider: ProviderFormat, models: Sequence[str]) -> list[
     return normalized_models
 
 
+def classify_provider_model_fetch_failure(
+    exception: Exception,
+) -> ProviderModelFetchFailure:
+    try:
+        import litellm
+    except Exception:
+        return ProviderModelFetchFailure.REQUEST_FAILED
+
+    if isinstance(exception, (litellm.APIConnectionError, litellm.Timeout)):
+        return ProviderModelFetchFailure.CONNECTION_FAILED
+    if isinstance(
+        exception, (litellm.AuthenticationError, litellm.PermissionDeniedError)
+    ):
+        return ProviderModelFetchFailure.ACCESS_DENIED
+    if isinstance(exception, litellm.RateLimitError):
+        return ProviderModelFetchFailure.RATE_LIMITED
+    if isinstance(
+        exception, (litellm.ServiceUnavailableError, litellm.InternalServerError)
+    ):
+        return ProviderModelFetchFailure.PROVIDER_UNAVAILABLE
+    if isinstance(exception, (ConnectionError, TimeoutError)):
+        return ProviderModelFetchFailure.CONNECTION_FAILED
+    return ProviderModelFetchFailure.REQUEST_FAILED
+
+
 def list_provider_models(
     *,
     provider: ProviderFormat,
@@ -261,12 +300,17 @@ def list_provider_models(
         configure_litellm_logging()
         model_lister = get_valid_models
 
-    models = model_lister(
-        api_base=normalize_provider_base_url(provider, base_url),
-        api_key=secret_reference,
-        check_provider_endpoint=True,
-        custom_llm_provider=provider_litellm_name(provider),
-    )
+    try:
+        models = model_lister(
+            api_base=normalize_provider_base_url(provider, base_url),
+            api_key=secret_reference,
+            check_provider_endpoint=True,
+            custom_llm_provider=provider_litellm_name(provider),
+        )
+    except Exception as exc:
+        raise ProviderModelFetchError(
+            classify_provider_model_fetch_failure(exc)
+        ) from exc
     return unique_model_names(provider, models)
 
 

package/backend/src/flowent/routes/providers.py

@@ -1,9 +1,25 @@
-from fastapi import FastAPI
+from fastapi import FastAPI, HTTPException, status
 
-from flowent.api_models import ProviderModelsRequest, ProviderModelsResponse
-from flowent.llm import list_provider_models
+from flowent.api_models import (
+    ProviderModelsFailureResponse,
+    ProviderModelsRequest,
+    ProviderModelsResponse,
+)
+from flowent.llm import (
+    ProviderModelFetchError,
+    ProviderModelFetchFailure,
+    list_provider_models,
+)
 from flowent.storage import StateStore, StoredProvider, StoredSettings
 
+PROVIDER_MODEL_FAILURE_STATUS: dict[ProviderModelFetchFailure, int] = {
+    ProviderModelFetchFailure.CONNECTION_FAILED: status.HTTP_502_BAD_GATEWAY,
+    ProviderModelFetchFailure.ACCESS_DENIED: status.HTTP_403_FORBIDDEN,
+    ProviderModelFetchFailure.RATE_LIMITED: status.HTTP_429_TOO_MANY_REQUESTS,
+    ProviderModelFetchFailure.PROVIDER_UNAVAILABLE: status.HTTP_503_SERVICE_UNAVAILABLE,
+    ProviderModelFetchFailure.REQUEST_FAILED: status.HTTP_400_BAD_REQUEST,
+}
+
 
 def register_provider_routes(app: FastAPI, *, store: StateStore) -> None:
     @app.post("/api/providers")
@@ -17,13 +33,20 @@ def register_provider_routes(app: FastAPI, *, store: StateStore) -> None:
 
     @app.post("/api/providers/models")
     async def provider_models(request: ProviderModelsRequest) -> ProviderModelsResponse:
-        return ProviderModelsResponse(
-            models=list_provider_models(
-                base_url=request.base_url,
-                provider=request.provider,
-                secret_reference=request.secret_reference,
-            ),
-        )
+        try:
+            return ProviderModelsResponse(
+                models=list_provider_models(
+                    base_url=request.base_url,
+                    provider=request.provider,
+                    secret_reference=request.secret_reference,
+                ),
+            )
+        except ProviderModelFetchError as exc:
+            detail = ProviderModelsFailureResponse(code=exc.failure.value)
+            raise HTTPException(
+                detail=detail.model_dump(),
+                status_code=PROVIDER_MODEL_FAILURE_STATUS[exc.failure],
+            ) from exc
 
     @app.put("/api/settings")
     async def save_settings(settings: StoredSettings) -> StoredSettings:

package/backend/src/flowent/routes/system.py

@@ -28,14 +28,13 @@ def register_system_routes(
     @app.get("/api/state")
     async def app_state() -> StoredState:
         state = state_with_current_model_context_window(store.read_state())
-        active_run = runtime.active_run()
+        active_response = runtime.current_response()
         update: dict[str, object] = {
-            "active_run_event_index": active_run.latest_event_index
-            if active_run
+            "is_responding": active_response is not None
+            and not active_response.is_done,
+            "response_event_index": active_response.latest_event_index
+            if active_response
             else 0,
-            "active_run_id": active_run.id
-            if active_run and not active_run.is_done
-            else None,
             "mcp_servers": mcp_manager.servers_with_status(state.mcp_servers),
             "skills": discover_skills(cwd, store),
         }

package/backend/src/flowent/routes/workspace.py

@@ -9,7 +9,6 @@ from flowent.api_models import (
     WorkspaceMessageEditResponse,
     WorkspaceMessagesRequest,
     WorkspaceRespondRequest,
-    WorkspaceRunResponse,
 )
 from flowent.logging import TRACE_LEVEL
 from flowent.storage import StateStore
@@ -42,45 +41,54 @@ def register_workspace_routes(
             len(request.content),
         )
         logger.log(TRACE_LEVEL, "Workspace edited user content=%r", request.content)
-        messages, run = runtime.edit_message(
+        messages, response = runtime.edit_message(
             message_id,
             action=request.action,
             content=request.content,
         )
         return WorkspaceMessageEditResponse(
+            is_responding=response is not None,
+            messages=messages,
+        )
+
+    @app.post("/api/workspace/messages/{message_id}/errors/{error_id}/retry")
+    async def retry_workspace_error(
+        message_id: str,
+        error_id: str,
+    ) -> WorkspaceMessageEditResponse:
+        logger.info(
+            "Workspace error retry requested message_id=%s error_id=%s",
+            message_id,
+            error_id,
+        )
+        messages, response = runtime.retry_error(
+            message_id,
+            error_id=error_id,
+        )
+        return WorkspaceMessageEditResponse(
+            is_responding=response is not None,
             messages=messages,
-            run_id=run.id if run else None,
         )
 
     @app.post("/api/workspace/clear")
     async def clear_workspace() -> WorkspaceClearResponse:
         messages = runtime.clear()
-        await runtime.notify_cleared_runs()
+        await runtime.notify_cleared_response()
         return WorkspaceClearResponse(messages=messages)
 
-    @app.post("/api/workspace/runs")
-    async def start_workspace_run(
-        request: WorkspaceRespondRequest,
-    ) -> WorkspaceRunResponse:
-        logger.info("Workspace run requested content_length=%s", len(request.content))
-        logger.log(TRACE_LEVEL, "Workspace user content=%r", request.content)
-        run = runtime.create_run(request.content, message_id=request.message_id)
-        return WorkspaceRunResponse(run_id=run.id)
-
-    @app.get("/api/workspace/runs/{run_id}/stream")
-    async def stream_workspace_run(
-        run_id: str,
+    @app.get("/api/workspace/stream")
+    async def stream_workspace_response(
         after: int = Query(default=0, ge=0),
     ) -> StreamingResponse:
-        run = runtime.run_by_id(run_id)
+        response = runtime.stream_current_response()
         return StreamingResponse(
-            runtime.run_stream(run, after),
+            runtime.response_stream(response, after),
             media_type="text/event-stream",
         )
 
-    @app.post("/api/workspace/runs/{run_id}/stop")
-    async def stop_workspace_run(run_id: str) -> dict[str, bool]:
-        runtime.stop_run(run_id)
+    @app.post("/api/workspace/stop")
+    async def stop_workspace_response() -> dict[str, bool]:
+        runtime.stop_response()
         return {"ok": True}
 
     @app.post("/api/workspace/compact", response_class=StreamingResponse)
@@ -98,8 +106,10 @@ def register_workspace_routes(
             "Workspace response requested content_length=%s", len(request.content)
         )
         logger.log(TRACE_LEVEL, "Workspace user content=%r", request.content)
-        run = runtime.create_run(request.content, message_id=request.message_id)
+        response = runtime.start_response(
+            request.content, message_id=request.message_id
+        )
         return StreamingResponse(
-            runtime.run_stream(run, include_snapshots=False),
+            runtime.response_stream(response, include_snapshots=False),
             media_type="text/event-stream",
         )

package/backend/src/flowent/state/models.py

@@ -217,6 +217,7 @@ class StoredMessage(BaseModel):
     status: str = Field(
         default="completed", exclude_if=lambda value: value == "completed"
     )
+    summary: str = Field(default="", exclude_if=lambda value: value == "")
     thinking: str = Field(default="", exclude_if=lambda value: value == "")
     tools: list[StoredToolItem] = Field(default_factory=list)
     usage_info: TokenUsageInfo | None = Field(
@@ -241,8 +242,8 @@ class StoredCompactionCheckpoint(BaseModel):
 class StoredState(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
-    active_run_event_index: int = 0
-    active_run_id: str | None = None
+    is_responding: bool = False
+    response_event_index: int = 0
     is_compacting: bool = False
     mcp_servers: list[StoredMcpServer]
     messages: list[StoredMessage]

package/backend/src/flowent/state/schema.py

@@ -83,6 +83,7 @@ def migrate(connection: sqlite3.Connection) -> None:
             id TEXT PRIMARY KEY,
             author TEXT NOT NULL,
             content TEXT NOT NULL,
+            summary TEXT NOT NULL DEFAULT '',
             status TEXT NOT NULL DEFAULT 'completed',
             usage_info TEXT,
             position INTEGER NOT NULL
@@ -154,6 +155,10 @@ def migrate(connection: sqlite3.Connection) -> None:
         )
     if "usage_info" not in message_columns:
         connection.execute("ALTER TABLE messages ADD COLUMN usage_info TEXT")
+    if "summary" not in message_columns:
+        connection.execute(
+            "ALTER TABLE messages ADD COLUMN summary TEXT NOT NULL DEFAULT ''"
+        )
     settings_columns = table_columns(connection, "settings")
     if "reasoning_effort" not in settings_columns:
         connection.execute(

package/backend/src/flowent/state/store.py

@@ -79,6 +79,7 @@ class StateStore:
                     ],
                     id=row["id"],
                     status=row["status"],
+                    summary=row["summary"],
                     thinking=row["thinking"],
                     tools=[
                         StoredToolItem.model_validate(tool)
@@ -90,7 +91,7 @@ class StateStore:
                 )
                 for row in connection.execute(
                     """
-                    SELECT id, author, content, tools, thinking, groups, status, usage_info
+                    SELECT id, author, content, summary, tools, thinking, groups, status, usage_info
                     FROM messages
                     ORDER BY position, id
                     """
@@ -576,6 +577,7 @@ class StateStore:
                     id,
                     author,
                     content,
+                    summary,
                     tools,
                     thinking,
                     groups,
@@ -583,13 +585,14 @@ class StateStore:
                     usage_info,
                     position
                 )
-                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
                 """,
                 [
                     (
                         message.id,
                         message.author,
                         message.content,
+                        message.summary,
                         json.dumps(
                             [
                                 tool.model_dump(exclude_none=True)
@@ -635,6 +638,7 @@ class StateStore:
                     id,
                     author,
                     content,
+                    summary,
                     tools,
                     thinking,
                     groups,
@@ -642,10 +646,11 @@ class StateStore:
                     usage_info,
                     position
                 )
-                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
                 ON CONFLICT(id) DO UPDATE SET
                     author = excluded.author,
                     content = excluded.content,
+                    summary = excluded.summary,
                     tools = excluded.tools,
                     thinking = excluded.thinking,
                     groups = excluded.groups,
@@ -657,6 +662,7 @@ class StateStore:
                     message.id,
                     message.author,
                     message.content,
+                    message.summary,
                     json.dumps(
                         [tool.model_dump(exclude_none=True) for tool in message.tools]
                     ),